• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Intel Cannonlake, Ice Lake, Tiger Lake & Sapphire rapid thread

Page 90 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

jpiniero

Diamond Member
Oct 1, 2010
7,279
895
126
Last article I had seen mentioning Hewitt Lake said it would be Cascade Lake-based, but if Intel has 10nm Xeon D ready to go, I'd be surprised if they bothered releasing it. They already have Skylake Xeon D.
Could just be like the rest of the Icelake lineup, where it's being covered by 14 nm products.
 

IntelUser2000

Elite Member
Oct 14, 2003
6,565
1,090
126
It could be seen as bad, only thing is sisoft and its as synthetic of a test as it can be, so the truth is we still don't know.

Until it launches no one knows. Leaks are only for specs, but when do we get to know how the specs perform? On release date!

Nice find @Dayman1225 for Gen 11. Improvements everywhere.

The Linley paper mentions that Gen 11 forgoes DP FP capability and reduces Int ops performance in half. Likely the reason why it reduces size by 25% at same process. Gen 9 has 1/4 DP FP. It doesn't matter though.

Gen 9 reduces EU size over Gen 8 by 16% eliminating the ability to use both AoS and SoA models to just SoA.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
6,565
1,090
126
Gen 11 likely uses immediate mode tiling similar to AMD and Nvidia rather than deferred mode.

The original Extreme Graphics, the second generation, and the GMA 9xx series used immediate mode tiling. They gave it up starting with the programmable shader based GMA X3000.

So really we've come full circle.

The loss of tiling, cutting down ROPs by half to 2, and very weak geometry engine is what caused the X3000 series to be often slower than the GMA 950 despite the former supporting full DX9 and programmable shaders.
 

jpiniero

Diamond Member
Oct 1, 2010
7,279
895
126
Iris Pro Graphics 950/40/30 and UHD Graphics 920/10, all GT2 configuration.
The 32/48/64 numbers at the end seem to indicate how many EUs it has however. They can always change the branding name later I guess, since it doesn't seem right to have GPUs with the same name and different EU counts.

Wonder what GT0.5 means... maybe no EUs activated or only like 8?
 

Dayman1225

Senior member
Aug 14, 2017
972
520
106
The 32/48/64 numbers at the end seem to indicate how many EUs it has however. They can always change the branding name later I guess, since it doesn't seem right to have GPUs with the same name and different EU counts.
Good eye, just notice Iris Plus 950 is 64EUs and 25w? Assuming that’s right
 

IntelUser2000

Elite Member
Oct 14, 2003
6,565
1,090
126
The Gen 11 whitepaper has interesting detail about the tile rendering its using.

It's entirely theirs, considering searching for it shows a patent dating back to 2015.

POSH - Position Only Tile Rendering
 

Dayman1225

Senior member
Aug 14, 2017
972
520
106
So Intel has a Data Centric Innovation day on the second of April where Cascade Lake will almost certainly be launched

I know this isn’t entirely relevant to this thread but I don’t think Cascade Lake needs a new thread created for it as it’s more or less Skylake SP refreshed with Optane DIMMs, AVX512 VNNI and Smeltdown security mitigation’s. Though I doubt Cascade Lake will be the only thing presented.
 

IntelUser2000

Elite Member
Oct 14, 2003
6,565
1,090
126
which makes it 20% improvement at the same clock not bad at all
There seems to be an improvement, but exact numbers are difficult to say.

With otherwise identical configuration, Geekbench scores can vary 50% or more. Compared to the top Windows scores, this one seems to be at 10% faster per clock.

I've seen earlier results that showed 15-20% improvement, but again, Geekbench is too synthetic. The biggest fault being it scales linearly with clock speed. Who knows if 10% in Geekbench means 5% elsewhere, or 20%?

You'll never get a proper comparison until the official launch date. It'll get more accurate, but never fully accurate until we see reviews from multiple sites and configurations.
 
  • Like
Reactions: coercitiv

epsilon84

Senior member
Aug 29, 2010
978
653
136
A Ryzen 1500x first generation 2 years old beats that, with the same number of cores and threads. Why is that impressive ? Or is that a laptop CPU ?
You answered your own question. 1500X turbos to 3.7GHz compared to 2.6GHz on this chip, its clearly not a desktop CPU. The fact that they are within the same ballpark despite a 40%+ clockspeed deficit is impressive, no?

Of course I'll take everything with a big grain of salt at this stage, and its only one bench.

I'm an enthusiast / overclocker at heart but of course I welcome more efficient laptop CPUs as it actually makes more of a difference to my work, my 2C/4T Surface Pro is seriously underpowered and constantly under high CPU usage, I'm forced to use my 4C/4T i5 laptop for any 'serious' work. If you can get 4C/8T level of performance on say a 15W power envelope then thats a massive thing for ultrabook / convertible tablet+laptop users.
 

TheGiant

Senior member
Jun 12, 2017
639
241
86
For Geekbench, you have to look at the individual tests themselves.

https://browser.geekbench.com/v4/cpu/compare/12411073?baseline=12472642

Compared to a 2.7 Ghz 6100H, it's like 5-10% faster in most of the tests outside of the AVX aware ones.
Well 6100H is a 35W TDP product, I am not that convinced it is a good comparison

Better a 6X00U, which looks like the same result at 3100 MHz so 3100/2600 so it looks like 20%

but as others said, better wait.....

even if its 10% at 4.5GHz it will be nearly like comparison between 4.5GHz CFL and 5GHz CFL and that is a huge difference

I am looking for a replacement for my SP4 PRO i5-6300U and currently nothing on the market that doesn't throttle in minutes in that form factor
 

ASK THE COMMUNITY

TRENDING THREADS